Read the original at HF Daily Papers
MAServeBench benchmarks multi-agent LLM serving systems using hierarchical collaboration workloads to evaluate hardware performance through configurable stages and semantic traces.
Carried by: HF Daily Papers. First seen: .